QasmModule Circuit Drawer

[arulandu]

Summary of changes

Added circuit drawing functionality for QasmModule. Closes #61.

[ryanhill1]

This progress looks good! A couple things:

1. Gate angle formatting: For parameterized gates, the angle below the gate's name goes all the way out to the edges of the gate box, and for negative numbers, it bleeds over the sides. I think remove the parentheses around the angle and making the font size of the angle slightly smaller than the gate name could make it fit better, and allow the main focus to be on the gate name itself.

2. Support for barriers: Add support for barriers. Currently being omitted from the diagram:

from pyqasm import loads

qasm3 = """
OPENQASM 3;
include "stdgates.inc";
qubit[2] q;
h q;
barrier q[0], q[1];
cnot q[0], q[1];
"""

module = loads(qasm3)

module.draw()

3. Measurement gate formatting:

• Measurement gates are currently being stacked on top of one another which makes it difficult to read. Would be better to space them out.
• The arrow pointing to the classical bit to measure currently goes over that bottom line.
• Let's make the arrow for the measure a little bit smaller
• The current color for the measurement gates is a bit dark. Maybe a slightly lighter grey?

Ours:

from pyqasm import loads

qasm3 = """
OPENQASM 3;
include "stdgates.inc";
qubit[2] q;
bit[2] b;
h q;
cnot q[0], q[1];
b = measure q;
"""

module = loads(qasm3)

module.draw()

vs. qiskit:

from qbraid import transpile

qiskit_circuit = transpile(qasm3, "qiskit")

qiskit_circuit.draw(output='mpl')

Note: I'm not saying to copy the way they've done it, but we can both it agree that theirs looks very clean. So we can perhaps take inspiration from the elements which make their diagrams visually appealing.

4. Optional to exclude idle wires: Can you add an option idle_wires of type bool which defaults to True, but if False, excludes any qubits and clbits that don't have any operation on them? For example:

from pyqasm import loads

qasm3 = """
OPENQASM 3;
include "stdgates.inc";
qubit[2] q;
bit[2] b;
rx(-pi/2) q[0];
"""

module = loads(qasm3)

module.draw()

module.draw(ide_wires=False)

5. Duplicate image from draw output: For some reason, when a call the .draw() method within my VS code notebook, it is "drawing" it twice. See images below. Is the same thing happening for you? See below:

6. Split lines for long diagrams: For circuit diagrams that are very long, we should at a certain point, split it onto two lines so that it is readable. Currently the font just gets smaller and smaller into oblivion

[ryanhill1]

Fix merge conflicts

[arulandu]

TODO: support QuantumPhaseStatement

[ryanhill1]

I think qbraid-cli might have a bug with the headers formatting, try setting qbraid-cli==0.9.4 in [testenv:headers] in tox.ini and run tox -e linters again and that should fix the header bugs + the file that looks like it has 2,000+ lines changed. If it doesn't, let me know.

Also, what's causing all of the different CI build workflows to fail?

[TheGupta2012]

Amazing work @arulandu ! Here are some minor comments -

1. Add qubit labels for multiple lines : For circuits extending to multiple lines, it would be nice to have qubit markers for the wires. They are currently being omitted whenever the program extends -

import pyqasm
test = """
OPENQASM 3.0;
include "stdgates.inc";
qubit[2] q;
gate custom a, b {
    h a;
    z a;
    z a;
    y a;
    x a;
    rx(0.5) a;
}
for int i in [0:2] {
    custom q;
}
"""
circ = pyqasm.loads(test)
circ.draw()

2. External gates should be drawn as is : Given that we provide support for external gates, we should not unroll them before drawing -

import pyqasm
test = """
OPENQASM 3.0;
include "stdgates.inc";
qubit[2] q;
gate custom a, b {
    h a;
    z a;
    z a;
    y a;
    x a;
    rx(-0.5) a;
}
custom q;

"""
circ = pyqasm.loads(test)
circ.unroll(external_gates = ["custom"])
circ.draw()

3. Support for reset : I believe reset ops are not supported right now. I guess you can open an issue for extending the drawer with resets or add them in this PR, whichever works

[TheGupta2012]

I think qbraid-cli might have a bug with the headers formatting, try setting qbraid-cli==0.9.4 in [testenv:headers] in tox.ini and run tox -e linters again and that should fix the header bugs + the file that looks like it has 2,000+ lines changed. If it doesn't, let me know.

Also, what's causing all of the different CI build workflows to fail?

@arulandu I think same applies for the ctrl modifier PR where we're seeing a very big change in the src/pyqasm/maps/gates.py

[TheGupta2012]

@arulandu you need to add qbraid in the test extra for pyqasm. This is why the tests are failing in the CI pipeline.

Moreover, since we use pyqasm as a dependency in the qBraid-SDK and transpiler, I'm not entirely sold on using it for randomized testing. Would be better to have a large enough dir with ~50 tests, where we can store pre-transpiled circuits and use them for testing

[arulandu]

@TheGupta2012
I omitted labels on 1. since this is what qiskit does. I can add them back them though. For external gates, until draw supports logical constructs like while/if/etc, we have to unroll inside the draw. I could support .draw(external=...), but in my opinion, it makes more sense to support this when our drawer does support logical constructs so that we don't need to unroll inside the draw. This feels like a separate issue on its own. For 3, I can open an issue and get to it soon

[ryanhill1]

@arulandu I added a couple basic tests to compare against images for you that should be sufficient for now.

I did find one bug though with how barriers are displayed when moments don't line up:

from pyqasm import draw

qasm3 = """
OPENQASM 3;
include "stdgates.inc";
qubit[2] q;
x q[0];
barrier q;
"""

draw(qasm3)

When applied across all qubits, the barrier should form a single vertical line, not be broken across qubits.

[TheGupta2012]

@arulandu I added a couple basic tests to compare against images for you that should be sufficient for now.

I did find one bug though with how barriers are displayed when moments don't line up:

from pyqasm import draw

qasm3 = """
OPENQASM 3;
include "stdgates.inc";
qubit[2] q;
x q[0];
barrier q;
"""

draw(qasm3)

When applied across all qubits, the barrier should form a single vertical line, not be broken across qubits.

I think this is because of how barriers are added in a moment. I feel this is expected draw behavior since we are unrolling the program before the draw. This will break down each barrier to single qubits and thus add each barrier statement to distinct moments. -

Possible solution

Add something like unroll_barriers = False param in unroll which will only do semantic check and leave them as is. Can be useful for draw so that we can find the qubits on which a barrier is applied. Once we have that, we can identify the qubit which has the max depth at the barrier application point and add barrier to that moment. Eg -

from pyqasm import draw

qasm3 = """
OPENQASM 3;
include "stdgates.inc";
qubit[2] q1;
x q1[0];
barrier q1;

qubit[4] q2;
h q2[1];
barrier q2;
h q2;
"""

draw(qasm3)

In above image , we will apply 2 barriers -

1. Barrier 1 - q2[0] -> q2[3] , q2[1] will have max depth (2) at the barrier application, so barrier is added to moment at depth 2
2. Barrier 2 - q1[0] -> q1[1], q1[0] will again have max depth (2) at the barrier application, so barrier is added to moment at depth 2

[codecov-commenter]

Codecov Report

❌ Patch coverage is 92.68293% with 21 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/pyqasm/printer.py	92.63%	21 Missing ⚠️

📢 Thoughts on this report? Let us know!